Abstract: Techniques for audio understanding using fixed language models are provided. In one aspect, a system for performing audio understanding tasks includes: a fixed text embedder for, on receipt of a prompt sequence having (e.g., from 0-10) demonstrations of an audio understanding task followed by a new question, converting the prompt sequence into text embeddings; a pretrained audio encoder for converting the prompt sequence into audio embeddings; and a fixed autoregressive language model for answering the new question using the text embeddings and the audio embeddings. A method for performing audio understanding tasks is also provided.

Abstract: A panel device including a substrate, a conductor pad, a turning wire, and a circuit board is provided. The substrate has a first surface and a second surface connected to the first surface while a normal direction of the second surface is different from a normal direction of the first surface. The conductor pad is disposed on the first surface of the substrate. The turning wire is disposed on the substrate and extends from the first surface to the second surface. The turning wire includes a wiring layer in contact with the conductor pad and a wire covering layer covering the wiring layer. The circuit board is bonded to and electrically connected to the wire covering layer. A manufacturing method of a panel device is also provided herein.

Abstract: A fault detection method for a rotating machine based on sparse time synchronous averaging is disclosed, and the method includes: collecting a vibration signal and a rotating frequency or rotating frequency pulse signal of the rotating machine and performing analog-to-digital conversion to obtain the vibration signal and rotating speed information by a sensor; according to the type and number of detection components in the rotating machine, constructing a component-aware comb vector g based on the vibration signal and rotating speed information, wherein the type includes the gear, rotor and bearing; constructing a quasi-time synchronous average vector w based on the component-aware comb vector g; constructing a sparse time synchronous averaging model F by using the quasi-time synchronous average vector w; solving the sparse time synchronous averaging model F with an optimization solution algorithm to obtain a sparse spectrum and a reconstruction time signal.

Abstract: A computer-implemented method for fine-grained referring expression comprehension is provided. The computer-implemented method includes receiving, at a processor, a textual expression and an image as inputs and executing, at the processor, fine-grained referring expression comprehension. The executing includes decomposing the textual expression into different textual modules, extracting visual regional proposals from the image, using language-guided graph neural networks to mine fine-grained object relations from the visual regional proposals and aggregating different matching similarities between the different textual modules and the fine-grained object relations.

Abstract: In some embodiments, the present disclosure relates to a device having a semiconductor substrate including a frontside and a backside. On the frontside of the semiconductor substrate are a first source/drain region and a second source/drain region. A gate electrode is arranged on the frontside of the semiconductor substrate and includes a horizontal portion, a first vertical portion, and a second vertical portion. The horizontal portion is arranged over the frontside of the semiconductor substrate and between the first and second source/drain regions. The first vertical portion extends from the frontside towards the backside of the semiconductor substrate and contacts the horizontal portion of the gate electrode structure. The second vertical portion extends from the frontside towards the backside of the semiconductor substrate, contacts the horizontal portion of the gate electrode structure, and is separated from the first vertical portion by a channel region of the substrate.

Abstract: In some embodiments, the present disclosure relates to a semiconductor device comprising a source and drain region arranged within a substrate. A conductive gate is disposed over a doped region of the substrate. A gate dielectric layer is disposed between the source region and the drain region and separates the conductive gate from the doped region. A bottommost surface of the gate dielectric layer is below a topmost surface of the substrate. First and second sidewall spacers are arranged along first and second sides of the conductive gate, respectively. An inner portion of the first sidewall spacer and an inner portion of the second sidewall spacer respectively cover a first and second top surface of the gate dielectric layer. A drain extension region and a source extension region respectively separate the drain region and the source region from the gate dielectric layer.

Abstract: An optical system may include a light source to provide a beam of light. The optical system may include a reflector to receive and redirect the beam of light. The optical system may include a light gate having an opening to permit the beam of light, from the reflector, to travel through the opening. The optical system may include a light sensor to receive a portion of the beam of light after the beam of light travels through the opening, and convert the portion of the beam of light to a signal. The optical system may include a processing device to determine whether a notch of a wafer is in an allowable position based on the signal.

Abstract: A management method for a cutting tool of a machine tool is provided. A user host imports a graphical data of a workpiece and a cutting tool database into CAM, and a user may select cutting tools to establish a complete machining process. T codes are replaced with electronic tags. Before executing a machine code comprising the electronic tags, the machine tool re-scans the cutting tools assembled at a tool turret and updates cutting tool arrangement information to ensure that the tool turret is equipped with the cutting tools used in the machining process. When the machine tool stores the electronic tags, a ready message is displayed for reminding an operator. When the electronic tags are not stored, the current storage location of a cutting tool lacked in the machining process or a warning message is displayed at the operation interface for reminding the operator.

Abstract: A path calculation method, apparatus, and device, to implement network slicing. The path calculation method includes: obtaining an algorithm constraint supported by each of a plurality of network devices, where the algorithm constraint is a constraint of a forwarding path algorithm, and the forwarding path algorithm is used to calculate, for the network device, a forwarding path that meets the algorithm constraint; performing network topology division on the plurality of network devices, where network devices in a same network topology support a same algorithm constraint; and calculating a forwarding path between network devices in each network topology based on the algorithm constraint, of the forwarding path algorithm, corresponding to the network topology.

Abstract: A screw compressor slide valve with gas pulsation attenuation function includes valve walls and internal through holes. A screw compressor includes the screw compressor slide valve. A part of exhausted gas of the screw compressor is directly discharged from an exhaust port to an exhaust chamber to form a main exhaust flow channel; another part of the exhausted gas is discharged from the exhaust port into the exhaust chamber after being delayed by the internal through holes, which forms a branched exhaust flow channel. Since the branched exhaust flow channel is longer than the main exhaust flow channel, when the gas pulsation in the branched exhaust flow channel lags behind that in the main exhaust flow channel by 180-degree in phase, two gas pulsations in the two flow channels are offset due to opposite gas pulsation phases, which attenuates the gas pulsation, thereby suppressing induced vibration and noise.

Abstract: An on-chip integrated circuit, a data processing device and a method are provided. The on-chip integrated circuit includes: a processor circuit and an accelerator circuit. The processor circuit includes a processor and a data storage area, the processor is connected to the data storage area through a first bus in the processor circuit. The accelerator circuit includes an accelerator and a second bus, the accelerator is connected to the second bus, and the second bus is bridged with the first bus corresponding to the data storage area, to perform data interaction between the accelerator and the data storage area, which can reduce the congestion on a bus of the processor and improve the quality of service of the application.

Abstract: A network edge computing method includes receiving, by an edge data node, a service request at least processed by network edge computation scheduling; and routing, according to a service port involved in the service request, the service request to a container of the edge data node, to be processed by the container.

Abstract: An array substrate (10) includes: a base plate; gate lines (11), data lines (12), and pixel units, all on the base plate, wherein each pixel unit includes a display electrode (13) having at least one electrode portion (131), each electrode portion (131) includes a main chain electrode (1311) and branch chain electrodes (1312) all electrically connected to the main chain electrode (1311); and a black matrix layer (14) on a side of the pixel units distal to the base plate, wherein the black matrix layer (14) includes openings (141) and a light shielding portion (142). On the base plate, an orthogonal projection of each opening (141) covers orthogonal projections of the branch chain electrodes (1312) of one corresponding display electrode (13), and an orthogonal projection of the light shielding portion (142) covers orthogonal projections of the gate lines (11), the data lines (12) and the main chain electrode (1311).

Abstract: The present disclosure relates to a display panel (10) and a method of manufacturing the same. The display panel (10) includes a substrate (101), a number of light transmissive units (320), and an package layer (600). The substrate (101) has a light transmissive region (700) and a pixel display region (800). The light transmissive region (700) and the pixel display region (800) are complementarily disposed with respect to the substrate (101). The plurality of light transmissive units (320) are disposed in the light transmissive region (700) and overlay a surface of the substrate (101) within the light transmissive region (700). The package layer (600) overlays a surface of a number of light transmissive units (320). In the present disclosure, the substrate (101) has a different structure in the light transmissive region (700) and the pixel display region (800).

Abstract: An electrode structure, a display panel, and an electronic device are provided, the electrode structure includes a first electrode portion, a second electrode portion and a conductive connection portion, the first electrode portion includes a first connection bar having a first side and a second side and a plurality of first electrode strips, ends of adjacent first electrode strips away from the first connection bar are open; the second electrode portion includes a second connection bar at a position of the first side away from the second side and a plurality of second electrode strips, the second connection bar includes a third side and a fourth side; the second electrode strips are connected with the second connection bar, ends of adjacent second electrode strips away from the second connection bar are open; ends of the conductive connection portion are connected with the first connection bar and the second connection bar.

Abstract: The present invention belongs to the field of treatment of urban organic wastes, and specifically relates to a high-value treatment system or method for urban wet garbage. According to the present invention, through the steps such as oil extraction, high-efficiency hydrolysis, high-value biological conversion, simultaneous recovery of released nitrogen and phosphorus and deep utilization of residues, urban wet garbage is converted into acetic acid by high-value treatment, produced by-products including carbon dioxide and hydrogen are biologically converted into acetic acid, released nitrogen and phosphorus are recycled into slow-release fertilizers, and solid residues are used to prepare materials capable of promoting conversion of the wet garbage into acetic acid through high-value treatment. According to the present invention, not only can high-value treatment of the urban wet garbage be realized, but also produced waste gases and waste residues are recycled.

Abstract: Embodiments of this application disclose a packet sending method, a device, and a system, so that a specified network device does not use a backup forwarding path to forward a packet, thereby reducing, to some extent, a technical problem such as network resource waste or network congestion caused by a loop problem. The method includes: A first network device obtains a first packet destined for a destination device; the first network device adds a first indication identifier to the first packet to generate a second packet, where the first indication identifier is used to indicate a second network device to avoid using a backup forwarding path from the second network device to the destination device to send the second packet to the destination device; and the first network device sends the second packet to the second network device by using a first forwarding path.

Abstract: This application discloses a forwarding entry generation method, a packet sending method, a network device, and a system, so that a specified network device cannot use a backup forwarding path to forward a packet, thereby reducing, to some extent, a technical problem such as network resource waste or network congestion caused by a loop problem. The method includes: A first network device obtains routing information advertised by a second network device; and the first network device determines that the routing information advertised by the second network device matches a summary route stored in the first network device, and generates, based on a segment identifier of the second network device, a forwarding entry corresponding to the summary route. The forwarding entry includes the segment identifier of the second network device, and a forwarding path corresponding to the forwarding entry passes through the second network device.

Abstract: A network edge computing method includes receiving, by an edge data node, a service request; and routing, by the edge data node and according to a virtual IP address and port information in the service request, the service request to one or more containers corresponding to a service, to be processed by the one or more containers.

Abstract: A network edge computing method includes receiving, by an edge data node, a service request at least processed by network edge computation scheduling; and routing, according to a service port involved in the service request, the service request to a container of the edge data node, to be processed by the container.